Amazon SageMaker

Developer Guide

What Is Amazon SageMaker?
How It Works
- Machine Learning with Amazon SageMaker
- Explore and Preprocess Data
- Model Training
  - Incremental Training
- Model Deployment
  - Hosting Services
  - Batch Transform
- Validating Models
- Programming Model
Getting Started
- Step 1: Setting Up
  - Step 1.1: Create an Account
  - Step 1.2: Create an S3 Bucket
- Step 2: Create an Amazon SageMaker Notebook Instance
  - Step 2.1: (Optional) Customize a Notebook Instance
- Step 3: Train and Deploy a Model
- Step 4: Clean up
- Step 5: Additional Considerations
Automatic Model Tuning
- How Hyperparameter Tuning Works
- Defining Objective Metrics
- Defining Hyperparameter Ranges
- Example: Hyperparameter Tuning Job
- Design Considerations
Using Notebook Instances
- Creating a Notebook Instance
- Accessing Notebook Instances
  - Limit Access to a Notebook Instance by IP Address
  - Connect to a Notebook Instance Through a VPC Interface Endpoint
- Using Example Notebooks
- Set the Notebook Kernel
- Installing External Libraries and Kernels in Notebook Instances
Using Built-in Algorithms
- Common Information
- BlazingText
  - Hyperparameters
  - Model Tuning
- DeepAR Forecasting
- Factorization Machines
- Image Classification Algorithm
- K-Means Algorithm
- K-Nearest Neighbors
- Latent Dirichlet Allocation (LDA)
- Linear Learner
- Neural Topic Model (NTM)
- Object2Vec
- Object Detection Algorithm
- Principal Component Analysis (PCA)
- Random Cut Forest
- Sequence to Sequence (seq2seq)
- XGBoost Algorithm
Using Your Own Algorithms
- Using Your Own Training Algorithms
- Using Your Own Inference Code
  - Using Your Own Inference Code (Hosting Services)
  - Using Your Own Inference Code (Batch Transform)
- Example: Using Your Own Algorithms
Automatically Scaling Amazon SageMaker Models
- Configure Automatic Scaling for a Variant
- Editing a Scaling Policy
- Deleting a Scaling Policy
- Load Testing
- Additional Considerations
Using TensorFlow
- TensorFlow Model Training Code
- Examples: Using Amazon SageMaker with TensorFlow
  - Example 1: Using the tf.estimator
Using Apache MXNet
- Apache MXNet Model Training Code
- Examples: Using Amazon SageMaker with Apache MXNet
  - Example 1: Using the Module API
Using Chainer
Using PyTorch
Using Apache Spark
- Example 1: Amazon SageMaker with Apache Spark
  - Custom Algorithms for Model Training and Hosting on Amazon SageMaker with Apache Spark
  - Using the SageMakerEstimator in a Spark Pipeline
- Additional Examples: Amazon SageMaker with Apache Spark
Amazon SageMaker Libraries
Authentication and Access Control
- Overview of Managing Access
- Using Identity-Based Policies (IAM Policies)
- Amazon SageMaker API Permissions Reference
- Amazon SageMaker Roles
Monitoring
- Monitoring with CloudWatch
- Logging with CloudWatch
- Logging Amazon SageMaker API Calls with AWS CloudTrail
Best Practices
- Deployment Best Practices
Security
- Notebook Instance Security
- Connect to Amazon SageMaker Through a VPC Interface Endpoint
- Protect Training Jobs by Using an Amazon Virtual Private Cloud
- Protect Endpoints by Using an Amazon Virtual Private Cloud
- Protect Data in Batch Transform Jobs by Using an Amazon Virtual Private Cloud
Limits and Supported Regions
API Reference
- Actions
  - Amazon SageMaker Service
  - Amazon SageMaker Runtime
    - InvokeEndpoint
- Data Types
  - Amazon SageMaker Service
  - Amazon SageMaker Runtime
- Common Errors
- Common Parameters
Document History
AWS Glossary

Logging Amazon SageMaker with Amazon CloudWatch

To help you debug your training jobs, endpoints, and notebook instance lifecycle configurations, anything an algorithm container, a model container, or a notebook instance lifecycle configuration sends to stdout or stderr is also sent to Amazon CloudWatch Logs. In addition to debugging, you can use these for progress analysis.

Logs

Log Group Name	Log Stream Name
`/aws/sagemaker/TrainingJobs`	`[training-job-name]/algo-[instance-number-in-cluster]-[epoch_timestamp]`
`/aws/sagemaker/Endpoints/[EndpointName]`	`[production-variant-name]/[instance-id]`
`/aws/sagemaker/NotebookInstances`	`[notebook-instance-name]/[LifecycleConfigHook]`
`/aws/sagemaker/TransformJobs`	`[transform-job-name]/[instance-id]-[epoch_timestamp]`
`/aws/sagemaker/TransformJobs`	`[transform-job-name]/[instance-id]-[epoch_timestamp]/data-log`

Note

The /aws/sagemaker/NotebookInstances log group is made when you create a Notebook Instance with a Lifecycle configuration. For more information, see Step 2.1: (Optional) Customize a Notebook Instance .

Warning Javascript is disabled or is unavailable in your browser.

To use the AWS Documentation, Javascript must be enabled. Please refer to your browser's Help pages for instructions.

Document Conventions

« Previous Next »

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.